2.1 Study Site

The study was conducted in Baoqing County, Heilongjiang Province, Northeast China (46°41′ N, 132°01′ E). The region is a typical area distributed with black soils, located in the center of the Sanjiang Plain. It has a cold temperate continental monsoon climate. The 30-year (1989–2019) mean annual air temperature is 4.6°C, and the annual rainfall is 550–600 mm, two-thirds of which falls between May and September.

This area was originally occupied by wetlands dominated by Phragmites australis and Deyeuxia angustifolia. Agricultural utilization at this site dates back to approximately 100 years, which generated croplands with a clear sequence of cultivation gradients spanning from 5 to 100 years and divergent quantity and availability of SOC (Li et al. 2022; Li, Hong, et al. 2024). Croplands with the same quantity but different availability of SOC were selected in the current study (Chen 2025). The selected fields have been continuously cultivated with maize for 30 years (High carbon availability, HCA) and 85 years (Low carbon availability, LCA) under typical and consistent practices. These management practices involve a traditional up-down ridge tillage pattern, with ploughing performed annually, a single application of synthetic fertilizers, and removal of all crop residues post-harvest. The crops are maize (Zea mays L.), which is traditionally sown in early May and harvested in October under rain-fed conditions.

2.2 Field Experiment Design

Four treatments were arranged in a complete randomized block design with four replicates in HCA and LCA croplands, and each of the plots measured 3.9 m × 4.5 m. The treatments included no N or phosphorus (P) fertilization as control (CK), N (NF), and P (PF) fertilizers applied singly and together (NP). Urea was applied at a rate of 160 kg N ha⁻¹, split between starter fertilizer and side-dressing at a ratio of 1:1. P fertilizer was applied as calcium superphosphate (75 kg P₂O₅ ha⁻¹). Potassium (K) sulfate (60 kg K₂O ha⁻¹) was applied as the basal fertilizer for each treatment.

The experimental period encompassed two maize-cropping seasons, from May 2018 to October 2019 and May 2019 to October 2019. Following local practices, the field plots were split into ridges and furrows at a distance of 65 cm by rotary tillage after harvest in autumn. On 11 May 2018 and 9 May 2019, starter fertilizers were evenly applied to the ridges at a depth of 10 cm, and maize seeds were sown at a 20-cm plant spacing. Side-dressing with urea occurred on 7 July 2018 and 10 July 2019 at the maize V8 growth stage. Mature maize was harvested on 4 October 2018 and 11 October 2019. All crop residues were removed from the plots followed by manual tillage. Other field management practices, such as weed and pest control, were conducted according to the local agronomic practices.

2.3 Determination of Soil N₂O Fluxes

The static chamber method was used to measure soil N₂O, nitric oxide (NO), and carbon dioxide (CO₂) fluxes across 2 years from 14 May to 14 November in 2018 and from 10 May to 2 November in 2019. A stainless steel base frame (65 cm × 20 cm × 20 cm) was inserted 10 cm into the soil. During gas collection, a stainless-steel sampling chamber (65 cm × 20 cm × 15 cm) was placed on the base sink of the frame and filled with water to ensure airtightness.

To focus on the impact of extreme summer rainfall, the period following ice formation during winter was not included in this study. Gas samples were collected twice a week during the maize growing season and every 5 days after maize harvest until the soil froze. This sampling frequency has been proven to adequately depict the pattern of N₂O fluxes in the study region, including peaks after N fertilization and rainfall (Chen et al. 2016). Multiple heavy rainfall events in August 2019 caused soil water supersaturation in the low-lying HCA cropland over 2 months, preventing access to the experimental field and sample collection. Gas sampling was conducted from 9:00 am to 12:00 pm to minimize diurnal variations. Four samples were withdrawn using syringes from the chamber headspace at 0, 10, 20, and 30 min after chamber closure and injected into a pre-evacuated 22-mL glass bottle. N₂O and CO₂ concentrations were measured using gas chromatograph (GC) (Agilent 7890, Agilent Technologies). Daily calibrations were performed with standard gases obtained from the National Institute of Metrology, China. The NO concentration was determined using an NO-NO₂-NO_X analyzer (model 42i, Thermo Fisher Scientific), as described in the Supporting Information.

The gas fluxes were calculated using linear fitting models of the gas concentration against the sampling time. Cumulative N₂O emissions were calculated by sequentially summing the fluxes using linear interpolation between interval sampling dates. The N-fertilizer-induced N₂O emission factors (EF) were calculated by dividing the difference in N₂O emissions between the fertilized treatment and the control (CK) by the total amount of applied fertilizer N.

2.4 Measurement of Environmental and Soil Parameters

Climatic parameters, including the daily air temperature (AT) and precipitation, were recorded at a meteorological station during the experimental period. Synchronized with gas sampling, the soil temperature (ST) and volumetric soil water content (SWC) at a depth of 5 cm were measured using a geothermometer and a time domain reflectometry probe, respectively. As soil bulk density (BD) did not vary significantly among treatments during the experimental period, soil moisture content was expressed as water-filled pore space (WFPS) with SWC and average BD.

Surface soil (0–20 cm) was collected once every two gas sampling campaigns. Three soil samples in each plot were combined into one composite sample and then were immediately returned to the laboratory and stored at 4°C for the subsequent analyses. Soil $$$ {\mathrm{NH}}_4^{+} $$$ and $$$ {\mathrm{NO}}_3^{-} $$$ were extracted by 2-M KCl in a 1:5 soil-solution ratio and analyzed using a continuous-flow autoanalyzer (San⁺⁺ System, Skalar Analytical BV). Extractable inorganic nitrogen (EIN) was calculated by the sum of $$$ {\mathrm{NH}}_4^{+} $$$ and $$$ {\mathrm{NO}}_3^{-} $$$. DOC was extracted by deionized water in a 1:10 soil-water ratio and determined by a TOC analyzer (vario TOC Cube, Elementar Analysensystemes GmbH). The absorbance of the DOC solution at 254 nm was determined and divided by the DOC concentration and recorded as the specific ultraviolet absorption (SUVA), which was used as an indicator of the aromatic degree of DOC (Weishaar et al. 2003). The seasonal dynamics of soil $$$ {\mathrm{NH}}_4^{+} $$$, $$$ {\mathrm{NO}}_3^{-} $$$, and DOC concentrations and SUVA of DOC are presented in Figure S1. Additionally, according to the method described by Xiao et al. (2018), we calculated the fluorescence index (FI, representing the relative contribution of microbial and plant sources to dissolved organic matter) and humification index (HIX, a descriptor of the degree of humification) to further characterize the variation in the chemical composition of DOC during the waterlogging periods in the wet year.

Before the commencement of the experiment, soil sampling (0–20 cm) was collected to determine the main physicochemical properties. Soil pH, BD, particle size, and total P and K content were presented in Table S1. The SOC and total nitrogen (TN) were analyzed with a CN analyzer (Vario Max CN, Elementar). Soil carbohydrates, as readily utilized fractions to microbes, were measured using anthrone colorimetric procedures. The mineral-associated organic matter (MAOM) and particulate organic matter (POM) were separated from bulk soil by density fractionation. The molecular composition of SOC was further assessed by solid-state ¹³C nuclear magnetic resonance (NMR) spectroscopy, as detailed in the Supporting Information, and the results are shown in Figure S2.

Although HCA and LCA croplands had the same quantity of SOC, the former contained more available components, as indicated by greater DOC and carbohydrates content, higher POM/MAOM ratios, and lower SUVA values (Table 1). In addition, ¹³C NMR spectroscopy indicated that the SOC of HCA cropland had a larger abundance of labile groups, including O-alkyl C and di-O-alkyl C but less aromatic C and phenolic C (Figure S2). The calculated indices of lability and aromaticity of SOC further verified the greater availability and biodegradability of SOC for the HCA cropland (Bahadori et al. 2021; Luan et al. 2020).

2.5 Quantification of N-Cycling Functional Gene Abundance

In order to explore the microbial driving mechanism of N₂O emissions as affected by SOC availability and extreme rainfall events, we selected soil samples collected from two croplands during the summer when N₂O fluxes peaked in both years and from HCA cropland before and after waterlogging in the wet year. The specific sampling date and results of each measurement can be found in the Supporting Information (Figure S3). To compare the differences between the two fields with distinct SOC availability levels more clearly, the results of each plot were represented by the averages across the 2-year period.

DNA was extracted from 1 g of freeze-dried soil using the FastDNA Spin Kit for Soil (MP Biomedicals, CA), and its concentration was determined using a NanoDrop spectrophotometer (ND-2000, Thermo Fisher Scientific). To estimate the abundance of microbes in the nitrification and denitrification processes, archaeal amoA (AOA), bacterial amoA (AOB), nirK, nirS, nosZI, and nosZII were determined by quantitative PCR (qPCR). The 16S rDNA and ITS genes were also quantified as proxies for the total bacterial and fungal communities using taxon-specific primers. Quantifications were conducted in a total reaction volume of 20 μL, including 10 μL of SYBR Green Master Mix (Q112-02, Vazyme), 1 μL of DNA template, 0.4 μL of each primer, and 8.2 μL of ddH₂O on real-time PCR system (LightCycle480II,384, Roche). The primer sequences and qPCR protocols are listed in Table S2. The gene copy numbers of the target genes per gram of dry soil were calculated according to standard curves constructed using serial plasmid dilutions of known amounts of plasmid DNA-containing standard samples.

2.6 Laboratory ¹⁵N Tracing Experiment

A paired ¹⁵N-labelled laboratory assay was conducted to quantify the N₂O production pathways affected by C availability and moisture regime. Soil samples were collected from two croplands that had not received N fertilizer prior to the commencement of the experiment. We designed 60% and 120% WFPS treatments to simulate normal and waterlogged moisture conditions, respectively, based on the variation in soil WFPS as observed in situ over the 2-year period. Two ¹⁵N treatments were designed in three replicates, of which either $$$ {\mathrm{NH}}_4^{+} $$$ or $$$ {\mathrm{NO}}_3^{-} $$$ was labeled with ¹⁵N at 10 atom%. For each replicate, fresh soil (20 g on an oven-dried basis) was sieved to < 2 mm, packed into incubation bottles, and compacted to the target height approximating the field soil BD. After a 24-h pre-incubation period at 25°C, a solution of ¹⁵NH₄NO₃ or NH₄¹⁵NO₃ was added to the 250 mL bottles at a rate of 50 mg N kg⁻¹. This rate is equivalent to 120 kg N ha⁻¹, similar to the fertilization rate in the field, so as to simulate the mineral N concentrations when the N₂O fluxes peaked in the normal rainfall year as well as under waterlogged conditions. Soils were adjusted to the target moisture content by weighing based on the mass water content calculated from the WFPS and incubated at 25°C for 3 days. Five groups of soil were incubated for destructive soil sampling and N₂O emission measurements. The soil moisture content was maintained constant by weighing during the incubation. Gas and soil samples were collected immediately (0 day) and 1, 2, and 3 days after NH₄NO₃ addition. Soil $$$ {\mathrm{NH}}_4^{+} $$$ and $$$ {\mathrm{NO}}_3^{-} $$$ were extracted to determine the concentrations and ¹⁵N abundances by a continuous-flow autoanalyzer and elemental analyzer-isotope ratio mass spectrometry system (2000 Delta V Advantage, Thermo Fisher) (EA-IRMS), respectively. After extraction, soils were washed with deionized water three times, oven-dried at 55°C, and sieved through 150 μm to determine the ¹⁵N abundances of organic N by EA-IRMS. Gas samples were simultaneously collected from the incubation bottles to measure N₂O flux rates. Before the first gas sample collection, an equal volume of fresh air was added to maintain the headspace pressure. At each sampling event, 40 mL of headspace gas was collected and injected into two pre-evacuated vials at 0 and 6 h post-bottle sealing for the analyses of N₂O concentration and isotopic composition by GC and IRMS (MAT 253, Thermo Fisher), respectively.

2.7 Calculations and Statistical Analyses

Significant differences in inorganic N content, DOC content, SUVA value, cumulative N₂O emissions, and microbial abundances among treatments and between the two croplands over 2 years were assessed using one-way analysis of variance (ANOVA) in SPSS 20.0. The significance level was set at p < 0.05. Data normality (Kolmogorov–Smirnov test) and homogeneity of variance (Levene's test) were checked, and the ln transformation was used, if necessary. The random forest analysis was utilized with the ‘rfPermute’ package in R 4.0.3 software to evaluate the main factors affecting N₂O fluxes and the significance of each predictor variable. Redundancy analyses (RDA) were performed to verify the relationship between response variables (N₂O fluxes and abundance of target genes) and explanatory variables (soil properties) using Canoco5 software. Structural equation models (SEM) were developed to evaluate the direct and indirect impacts of edaphic conditions on N₂O fluxes during the waterlogging period using AMOS 22.0. Explanatory factors were selected for analysis in the SEM model based on Pearson, random forest, and RDA results. Path models were deemed acceptable if they met the criteria of goodness-fit-index (GFI) > 0.90, χ²/df < 3 and chi-squared-test (p) > 0.05.

3.1 Environmental Conditions

The ST varied following the trend of AT (Figure 1a) and exhibited no difference between the HCA and LCA croplands. Annual precipitation was 672 mm in 2018 and reached 899 mm in 2019, with the latter representing a historical maximum and a 77% increase compared with the long-term mean. Notably, the monthly precipitation in August 2019 reached a record-breaking high of 388 mm due to five rainstorm events (daily rainfall > 50 mm) (Figure 1b), resulting in supersaturated moisture conditions for the low-lying HCA cropland over 2 months. During the normal rainfall periods, WFPS varied from 27% to 77% until August 2019, when the soil was waterlogged, with WFPS remaining at 97%–120%.

3.2 Soil Mineral N and DOC

High $$$ {\mathrm{NH}}_4^{+} $$$ concentration was observed during the waterlogging period, even comparable to the peaks after side-dressing; while the $$$ {\mathrm{NO}}_3^{-} $$$ concentration was relatively low (Figure S1a,b). Mean $$$ {\mathrm{NH}}_4^{+} $$$ and $$$ {\mathrm{NO}}_3^{-} $$$ concentrations over the whole experimental period were significantly higher in HCA than LCA cropland without N fertilization while HCA had significantly lower $$$ {\mathrm{NH}}_4^{+} $$$ but higher $$$ {\mathrm{NO}}_3^{-} $$$ concentration when N was applied (Figure S4a,b). Moreover, the $$$ {\mathrm{NO}}_3^{-} $$$/$$$ {\mathrm{NH}}_4^{+} $$$ ratio was higher in HCA than LCA cropland (averaged at 5.8 vs. 3.4), potentially indicating a greater capacity of nitrification of HCA versus LCA (Figure S5).

DOC concentration over 2 years was significantly higher in HCA than the LCA cropland, averaged at 53.9 and 40.1 mg C kg⁻¹, respectively (Figure S4c), while the opposite was true for the SUVA value, indicating the greater biodegradability of DOC in the HCA soils (Figure S4d). In addition, significant stimulating effects of N and P fertilization on DOC concentrations were found in HCA soils rather than in LCA soils (Figure S4c).

3.3 Soil N₂O Emissions

During the normal rainfall year, seasonal dynamics of N₂O fluxes in HCA cropland were similar to LCA, and the peak value of 372 μg N m⁻² h⁻¹ was observed in the NF treatment of HCA (Figure 2). During the wet year, N₂O fluxes in HCA cropland peaked at 521 μg N m⁻² h⁻¹ during the waterlogging period, which were greatly higher than that in LCA (211 μg N m⁻² h⁻¹). It should be noted that the NO fluxes and N₂O/NO flux ratios remained at rather low levels under the waterlogging conditions (Figures S6 and S7).

During the normal rainfall year, the cumulative background N₂O emissions were significantly higher in HCA than LCA (0.72 vs. 0.26 kg N ha⁻¹) (Figure 3a). P fertilization showed no significant effect on N₂O emissions, while N fertilization from HCA and LCA croplands significantly increased them to 2.25 and 1.40 kg N ha⁻¹, respectively. During the record wet year, cumulative N₂O emissions became greatly higher in HCA than LCA, with background emissions of 4.38 and 0.35 kg N ha⁻¹, respectively, and 8.86 and 1.81 kg N ha⁻¹ for NF treatment (Figure 3b). Compared with the normal rainfall year, the N₂O-EF in the NF treatment significantly increased in the wet year, rising from 0.77% to 2.24% in the HCA cropland and from 0.57% to 0.73% in the LCA (Figure S8). Notably, the disparity in N₂O emissions between the two croplands was amplified by extreme rainfall events. The emission ratio of NO/N₂O in LCA was higher than that in HCA, especially in the wet year, which was no more than 0.15 for all treatments in HCA (Figure 3c,d).

3.4 N-Cycling Functional Genes

The abundance of AOA in the LCA croplands was greatly higher than that in the HCA croplands (Figure 4a), while that of AOB was higher in the HCA croplands, especially with N fertilization (Figure 4b). The nitrite-reducing community was dominated by nirK- rather than nirS-harbouring bacteria, and the abundance of nirK and nirS was significantly higher in the HCA cropland compared to the LCA cropland (Figure 4c,d). For the N₂O reducing community, nosZI-harbouring bacteria were approximately 100-fold more abundant than nosZII. Interestingly, the abundance of nosZI was higher in HCA croplands than in LCA croplands (Figure 4e), whereas nosZII, with stronger N₂O reduction potential, was lower in HCA cropland (Figure 4f). Additionally, the ratios of (nirS + nirK)/amoA and (nirS + nirK)/(nosZI + nosZII) were higher in the HCA croplands than in the LCA croplands (Figure 4g,h).

The abundance of N-cycling genes in the HCA cropland before and during the waterlogging period was determined to explore the microbial mechanisms underlying explosive N₂O fluxes (Figure S9). No significant changes were observed in the abundance of AOA, whereas AOB abundance decreased after the soil was waterlogged. The abundance of nirS was increased, whereas nirK decreased due to waterlogging conditions. The abundances of nosZI and nosZII decreased, particularly after N-fertilization treatment. Additionally, the abundance of nrfA was decreased. Changes in the abundance of 16S rDNA genes were minor, whereas ITS increased significantly after the soil was waterlogged.

3.5 Factors Driving N₂O Emissions

Random forest analysis revealed that the impacts of available C and N substrates on N₂O fluxes were greater than those of micro-environmental indicators such as moisture and temperature (Figure 5a–c). During the normal rainfall year, EIN and DOC were the most important variables driving N₂O fluxes. In the wet year, the primary drivers shifted towards DOC, EIN, and WFPS. An analysis which combined all data over the 2 years revealed that DOC remained the most critical factor in regulating N₂O fluxes, indicating that increased moisture reinforced the stimulatory effect of DOC on N₂O fluxes.

SEM explained 87% of the variability in N₂O fluxes (Figure 5d), which were directly related to the abundances of nirS, nirK, nosZI, and AOB/AOA ratios, indirectly regulated by DOC concentrations as well as directly related to the supply of EIN.

RDA analysis revealed that N₂O fluxes were positively correlated with DOC concentrations and AOB/AOA, (nirS + nirK)/amoA, and (nirS + nirK)/(nosZI + nosZII) ratios but negatively correlated with SUVA and nosZII. Furthermore, the abundances of nirS and nirK were positively linked to DOC concentrations (Figure 5e). Overall, the first axis explained 74.3% of the variability in total N₂O fluxes and microbial communities.

During the waterlogging period, there was a significantly positive correlation between the N₂O and CO₂ fluxes (Figure 6a). Furthermore, N₂O fluxes increased with increasing $$$ {\mathrm{NO}}_3^{-} $$$ concentration but decreased logarithmically with increasing the HIX of DOC (Figure 6b,c).

SEM was further constructed to investigate the cascading effects of soil moisture on N-related functional genes and N₂O fluxes as induced by extreme hydrological events (Figure 6d). The SEM explained 81% of the variability in N₂O fluxes, which were directly related to the abundances of nirS and nosZII, as well as WFPS and $$$ {\mathrm{NO}}_3^{-} $$$ supply. The N₂O fluxes, which were evidenced to be mainly influenced by nirS, were directly related to nirS and indirectly related to the DOC concentration. N₂O fluxes were directly negatively correlated with nosZII, which was mainly regulated by $$$ {\mathrm{NO}}_3^{-} $$$ directly.

3.6 N₂O Production Processes

Total N₂O emissions over the whole incubation period were 10.4 and 2.84 μg N kg⁻¹, respectively, from the HCA and LCA soils under 60%WFPS (Figure 7a). Elevated soil N₂O emissions were measured in both soils at higher moisture condition of 120%WFPS, with the emissions measured at 4971 and 1056 μg N kg⁻¹, respectively. The disparity in N₂O emissions between the two croplands was amplified under higher moisture conditions, which was consistent with the results of the in situ experiments.

Autotrophic nitrification was the primary process of N₂O production in HCA soils, with a higher contribution compared to LCA soils (55% vs. 33%), while the fraction of N₂O derived from heterotrophic nitrification was higher in LCA soils (26% vs. 48%) (Figure 7b). The contribution of denitrification to N₂O production was relatively low and insignificant between the two soils at 60%WFPS. On the contrary, the contribution of denitrification in HCA was significantly higher under 120%WFPS (97% vs. 92%), which was the predominant process of N₂O production. In addition, almost no N₂O was produced via autotrophic nitrification under extremely high moisture conditions.

4.1 Microbial Mechanism Driving N₂O Emissions as Regulated by Carbon Availability

N₂O emissions from HCA cropland were larger than those from LCA croplands during the normal rainfall year, with a significantly higher EF of 0.77% versus 0.57%. These values were lower than the IPCC default value of 1.00% and the global mean of 1.02% for maize croplands (Cui et al. 2021). Consistently, Chen et al. (2014) reported a value of 0.34% also in black soils under maize cultivation in Northeast China. We further synthesized the literature results of in situ N₂O emissions over at least the entire growing season from the black soil area located in Northeast China (Table S3, Figure S10). Out of expectation, cumulative N₂O emissions decreased with increasing SOC content, which was consistent with the negative correlation between them under conservation tillage (de Figueiredo et al. 2018) but contrary to the well-acknowledged positive relationships observed on regional or global scales (Bouwman et al. 2002; Cui et al. 2021; Ge et al. 2024). Moreover, the ratio of SOC to TN (C/N) exhibited a stronger relationship with N₂O emissions than SOC content, indicating that N₂O emissions were more tightly linked to the availability of SOC than its quantity (Yao et al. 2022). Although black soils have the highest SOC content among croplands worldwide, they are less enriched in active components (e.g., DOC and carbohydrates) (Li et al. 2022). Generally, a higher C/N ratio was associated with a larger proportion of recalcitrant organic C (e.g., aryl and phenolic groups; Figure S2), exhibiting slower decomposition and turnover rates (Soldatova et al. 2024), leading to a lower available substrate supply for denitrification and N₂O production (Chen, Li, et al. 2021). POM comprised of plant-derived materials has been shown to be positively associated with hotspots of microbial activity in the soils, including denitrification, despite MAOM being the main portion of bulk soil organic matter (Stuchiner et al. 2024). Greater soluble organic C could be derived from POM-fueled microbial processes and may create favorable condition for denitrification (Chen, Li, et al. 2021).

Indeed, the N₂O fluxes were observed to increase with higher DOC content, alongside a negative relationship with the aromaticity of DOC (Figure 5e). Similarly, Saari et al. (2009) also identified DOC concentration and decomposability as important controllers for N₂O release from forest soils. Previous research had indicated that DOC regulated N₂O emissions through directly affecting dominant microbial production pathways (Wieder et al. 2011; Yin et al. 2017). As shown in Figure 3 and Figure S7, the NO/N₂O ratio in the HCA cropland was well lower than 1, which indicated that more in situ N₂O was produced from heterotrophic denitrification, especially after rainfall and N-fertilization (Chen, Tu, et al. 2021; Ji et al. 2020). The DOC with high bioavailability serves as an electron donor and promotes denitrification to produce N₂O under lower O₂ conditions (Stuchiner and Von Fischer 2022). The nir(S + K)/amoA ratio, regarded as a valuable proxy for coupling the microbial genetic potential of denitrification versus nitrification (Liao et al. 2021), was positively related to N₂O fluxes and was significantly higher in HCA cropland (Figures 4g and 5e). Moreover, the SEM results confirmed that DOC stimulated N₂O production by facilitating the nirS- and nirK-type denitrifying bacteria (Figure 5d). It has been reported that these microbes prefer to inhabit soils rich in DOC and exhibit positive responses to labile C amendments (Florio et al. 2019). Therefore, the larger N₂O emissions in the HCA cropland were attributed to greater C availability, which promoted denitrification by fueling nirS- and nirK- denitrifiers.

Some denitrifiers carry the nosZ gene encoding N₂O reductase, which is the only known biological sink of N₂O. Thus, the determination of N₂O emissions by denitrification depends on the extent of reduction of N₂O to N₂ by the denitrifying community. The microbial ratio of (nirS + nirK)/(nosZI + nosZII) in the HCA cropland was higher than that in the LCA (Figure 4h), indicating a greater potential for N₂O production than further reduction (Liao et al. 2021). Indeed, the abundance of nosZI in HCA cropland was higher than that in LCA and had a positive correlation with N₂O fluxes (Figure 5e). In contrast, the abundance of nosZII, which has a stronger N₂O reduction ability (Linton et al. 2020), and the ratio of nosZII/(nirS + nirK) were lower in HCA than in LCA (Figure S11). Domeignoz-Horta et al. (2017) found that the diversity of nosZII but not of nosZI clade was the key determinant explaining the variation of in situ N₂O emissions. Here, we observed that nosZI-bacteria was positively dependent on the DOC supply. Many of the alphaproteobacterial species with nosZ clade I are plant-associated, including Bradyrhizobium spp. and Paracoccus spp., suggesting that N₂O-reducing organisms within these lineages are adapted to utilize a variety of labile C sources in root exudation, containing low molecular weight carbohydrates and organic acids (Abdelfattah et al. 2021; Maheshwari et al. 2023). In this study, the HCA cropland was characterized by higher crop yields with well-developed root systems (Table S4), which facilitated the exudation of easily decomposable substrates, thus enhancing the activity of nosZI rather than nosZII (Ai et al. 2020). Thus, the lower N₂O reduction potential in the HCA cropland was due to the poor abundance of nosZII. It should be noted that the responses of nosZII abundance and physiological activity to C substrates are still unclear. A deeper insight into the mechanisms driving niche differentiation between microbes with clades I and II nosZ is needed, especially in response to agricultural C management and climate change, thereby ultimately favoring the development of N₂O mitigation strategy (Hallin et al. 2018; Hiis et al. 2024; Xu et al. 2020).

Apart from the greater capacity for N₂O production via denitrification in HCA soils, stronger nitrification could also contribute to larger emissions. Actually, nitrification plays a dual role in the production of N₂O, by directly producing N₂O and providing $$$ {\mathrm{NO}}_3^{-} $$$ substrate for denitrification. A higher $$$ {\mathrm{NO}}_3^{-} $$$/$$$ {\mathrm{NH}}_4^{+} $$$ ratio of HCA than LCA potentially indicated its greater capacity for nitrification (Figure S5). Identically, the ¹⁵N tracing experiment showed that the contribution of autotrophic nitrification to N₂O production was higher in HCA soil (Figure 7b). Additionally, N₂O fluxes were increased with a higher ratio of AOB/AOA, and AOB was more abundant in HCA, as facilitated by its higher $$$ {\mathrm{NH}}_4^{+} $$$ supply compared to LCA (Figure S4a). Although ammonia-oxidizers do not use organic C as an energy source, higher-quality soil organic matter could promote microbial mining for N, (Liu et al. 2022) as evidenced by the higher gross rate of mineralization of HCA soils (Figure S12). This alleviated the limitation of substrate availability for nitrification by AOB and resulted in a more rapid and tight mineralization-nitrification coupling process (Figure S12) (Xiao et al. 2019), which in turn could provide more $$$ {\mathrm{NO}}_3^{-} $$$ substrate to stimulate the denitrification rates and thus N₂O production. Moreover, the N₂O production capacity of AOB was twice that of AOA, where production is limited to presumably abiotic hybrid formation (Hink et al. 2018; Prosser et al. 2020). Furthermore, the ability to denitrify is a common trait among AOB, so N₂O can be produced not only as a by-product of hydroxylamine metabolism but also likely through nitrifier denitrification as discussed above (Wrage-Mönnig et al. 2018).

In contrast, the fraction of N₂O produced via heterotrophic nitrification in the LCA soil was 1.4 times greater than that of autotrophic nitrification (Figure 7b). It has been traditionally deemed that heterotrophic nitrification principally occurs in grassland and forest soils, and is generally negligible in croplands (Zhang et al. 2015). However, Chen et al. (2014) observed a distinct occurrence of heterotrophic nitrification in cultivated black soils, which contributed up to 25%–49% of N₂O production. Fungi are considered the most efficient microbes performing heterotrophic nitrification (Martikainen 2022). Compared with bacteria, fungi are more acid-tolerant and require less N per unit biomass C accumulation, and ligninolytic fungal genera among Basidiomycetes and Ascomycetes could decompose recalcitrant compounds more efficiently (Wang and Kuzyakov 2024). Consequently, heterotrophic nitrification may play a vital role in N₂O production in LCA soils with more aromatic components (Chen et al. 2015; Zhang et al. 2019). However, the ability of N₂O production by fungal heterotrophic nitrification is much weaker compared to bacterial autotrophic nitrification (Zhang et al. 2015) and accordingly did not lead to high N₂O emissions from LCA soils. To get a more reliable perspective on the contribution of heterotrophs to nitrification, the physiology of soil heterotrophic nitrifiers and the trade-off between autotrophic and heterotrophic nitrification, as controlled by the dynamics of SOC quality, require further study.

4.2 Processes Responsible for N₂O Pulses Over the Waterlogging Period

During field monitoring, the study site experienced multiple heavy rainfall events in the summer of 2019, which became a record-breaking wet year with annual precipitation 1.8-fold higher than the long-term mean. Compared with the normal rainfall year, N₂O emissions from the HCA and LCA increased by 84% and 27%, respectively. In particular, the N₂O-EF of HCA cropland was greatly increased from 0.77% to 2.24%, which was three times greater than that of LCA and doubled the average global upland (1.05%) as estimated by Wang et al. (2020). The occurrence of climate extremes in the middle-high latitude regions is predicted to increase as accompanied by climate warming (Ombadi et al. 2023). Our results clearly indicated that extreme rainfall could potentially induce this region to become a “hotpot” of N₂O emissions, particularly for fields characterized by high SOC availability.

Enhanced N₂O emissions from HCA cropland during the extreme rainfall year were primarily due to explosive fluxes during the waterlogging period. Similar results were observed in cultivated black soils (Chen et al. 2016), semiarid grasslands (Li et al. 2023), and forests (Leitner et al. 2016). Nevertheless, the mechanisms responsible for emission pulses during the waterlogging period remain obscure because the transient and intense effects of extreme rainfall are highly challenging to capture (Fu et al. 2023; Knapp et al. 2017).

Montoya et al. (2022) supposed that in situ N₂O pulses after irrigation were ascribed to stimulated denitrification. Our laboratory ¹⁵N tracing experiment evidently revealed that larger N₂O emissions under supersaturated moisture conditions were almost exclusively produced by denitrification (Figure 7b). The flux ratio of NO/N₂O was no more than 0.12 during the waterlogging period (Figure S7b). This also suggested that the peak N₂O fluxes presented in the field could be predominantly attributed to denitrification when the soil moisture began to slowly decrease but remained saturated. It should be noted that the N₂O production pathways were quantified using ¹⁵N tracing incubations, which may have biases due to the differences between laboratory and field conditions, although they are unlikely to affect the main conclusions. The in situ ¹⁵N labeling method is recommended to decipher the processes of N₂O production in response to extreme hydrological events.

The significant relationship between CO₂ and N₂O fluxes during the waterlogging period suggested that N₂O emissions were associated with the activity of heterotrophic microbes, which were stimulated by soil extractable N and DOC (Figure 6a). The sudden increase in water content caused aggregates to disintegrate or organic–inorganic complexes to break down, leading to the dissolution of readily decomposable components to increase the DOC content (Védère et al. 2022). However, this increment did not immediately induce the peak of N₂O fluxes. Interestingly, we found that N₂O fluxes increased linearly with the HIX index, suggesting that the N₂O production rate was more likely to be regulated by DOC quality. The decreased values of SUVA, HIX, and FI indices (Figure S13) implied a reduction in the high-molecular-weight humic substances but possibly an increment in the plant-derived components during the waterlogging period (Zhou et al. 2019). In general, plants enhanced their resistance to anaerobic stress by releasing root exudation (Chai and Schachtman 2022; Henry et al. 2007). Additionally, we observed an increase in fungal abundance during the flooding process (Figure S9h). Compared to bacteria, fungi could more efficiently promote water transport and acquire resources through their mycelial networks, thus tending to dominate more of the activity of microbial communities in waterlogged environments (Wang et al. 2023). Fungi have a greater ability to decompose recalcitrant humic substances into low-molecular-weight molecules (Yuste et al. 2011), which were more easily utilized by heterotrophic denitrifiers and contributed to the enhanced N₂O emissions under extreme hydrological conditions.

The N₂O fluxes were observed to increase logarithmically with increasing $$$ {\mathrm{NO}}_3^{-} $$$ concentration during the waterlogging period (Figure 6b). However, we observed that an extremely lower concentration of $$$ {\mathrm{NO}}_3^{-} $$$ initially after the soil was saturated with the successive rainstorm events (< 2 mg N kg⁻¹), which was below the reported threshold of 3 mg N kg⁻¹ for high N₂O fluxes larger than 100 μg N m⁻² h⁻¹ (Chen et al. 2016). This was mainly because rapid flow paths might occur after the extreme rainfall, causing a “hot moment” for $$$ {\mathrm{NO}}_3^{-} $$$ leaching (Castellano et al. 2010). Additionally, Chen et al. (2016) deduced that the low $$$ {\mathrm{NO}}_3^{-} $$$ concentration after soil waterlogged might also be attributed to the occurrence of dissimilatory nitrate reduction to ammonium (DNRA), as supported by the simultaneous increase in $$$ {\mathrm{NH}}_4^{+} $$$ concentration. However, in the current study, the functional gene of nrfA did not exhibit a significant increase (Figure S9i), which suggested that high $$$ {\mathrm{NH}}_4^{+} $$$ might not be produced by the DNRA process but by the decomposition of organic matters under the waterlogging conditions (Qu et al. 2014). Subsequently, accompanied by the disappearance of soil surface water, nitrification gradually proceeded as soil aeration enhanced. A positive relationship between the abundance of AOA and $$$ {\mathrm{NO}}_3^{-} $$$ content was observed (Figure S14) possibly due to the tolerance of AOA communities to low O₂ concentrations and their preference for survival in stressful environments compared to AOB (Prosser et al. 2020). Therefore, the nitrification process driven by AOA in waterlogged soil might be the source of $$$ {\mathrm{NO}}_3^{-} $$$ as the substrate for N₂O production by denitrification. It is suggested that the dynamics of soil O₂ concentrations should be included and linked to in situ N₂O fluxes in future studies in order to elucidate the mechanisms of N₂O emissions, especially in rainfed croplands, where redox conditions are sensitive to rainfall events.

Apart from the enhancement of C availability and $$$ {\mathrm{NO}}_3^{-} $$$ supply, substantial post-rainfall N₂O emissions could also be related to the increase in denitrifying microbes (Montoya et al. 2022). In the present study, the abundance of nirS increased significantly, while nirK and both clades of nosZ decreased after soil waterlogged (Figure S9). Compared with nirK, nirS denitrifiers are better adapted to soil conditions with high moisture and low O₂ content (Qin et al. 2018; Shi et al. 2021). The nirS communities have been demonstrated to be more closely related to the supply of labile C substrates under sub-anoxic than aerobic conditions (Yuan et al. 2012), which implied distinct regulatory mechanisms of substrates on denitrifying microorganisms as affected by changing water conditions. Accordingly, enhanced C availability might have stimulated the activity of nirS denitrifiers during the waterlogging period in this study. As for the decreased abundance of nosZ gene, the inhibition of nosZ activity by increased $$$ {\mathrm{NO}}_3^{-} $$$ content might be the reason (Friedl et al. 2020; Hallin et al. 2018). This could be partly supported by the lower abundance of nosZ gene in soils treated with N-fertilization as compared with the CK treatments (Figure S15). Furthermore, N₂O fluxes during this period were observed to positively increased with the ratio of nirS/nosZII (Figure S16). Therefore, it could be concluded that the explosive N₂O emissions as induced by the extreme hydrologic conditions were mostly derived from the nirS-driven incomplete denitrification as promoted by the increased substrates of available C and $$$ {\mathrm{NO}}_3^{-} $$$.